Reliability Score: A Computational Evaluation of Trustworthiness in Web-Based Articles

Blarr, Audrey

Reliability Score: A Computational Evaluation of Trustworthiness in Web-Based Articles

Persistent URL

https://hdl.handle.net/10456/59285

Author(s)

Blarr, Audrey

Date Issued

May 7, 2025

Abstract

In a research world full of detrimental biases and the spread of inaccurate and outdated information, compiling a list of resources to utilize in personal research projects while fully trusting their accuracy can pose many difficulties. In recent times, where the spread of information online is extremely vast and AI-generations are increasingly realistic and deceitful, it can be easy for young and inexperienced researchers especially to label information they come across as factual without feeling obligated to do further, more intricate research into how and where exactly the information is being derived from. It’s also crucial for researchers to know which elements of a research article should be analyzed to determine the overall accuracy and relevancy of the information, which isn’t exactly common knowledge. This project aims to create a database through Python-style coding, in which young researchers can take any HTML research article’s URL and pass it through web scraping to derive a list of crucial information about the article, including the author, publisher, and published and/or modified dates, all necessary in determining reliability. Once the author and publisher elements are extracted, the values of each element will be passed into OpenAI's "chatgpt-4o-latest", a Large-Language Model (LLM) that utilizes AI to respond to prompts from the user. Along with resources provided by a Google API query on each topic, this model will conduct outside research on the author to determine their educational levels, expertise levels based on other research conducted, and any controversies they've been involved with, along with any controversies the publisher has been involved with. As these elements are compiled, their values will be mapped to numeric values that either add or subtract to a reliability score for that element, then each individual element's score will be averaged together to display a final reliability score, in hopes of providing a clear and concise answer for the researcher as to whether or not this article should be considered factual for their own research purposes. The goal of this project is to determine which elements of a research article should be taken into consideration while evaluating reliability of the provided information, and how exactly those elements’ values can be mapped to a concise reliability score. Opportunities will also be posed in which the database’s ability to create an accurate scoring of any research article can be tested, and any changes necessary to improve the accuracy of this system will be stated as means of future research and implementation.

Major

Informatics

First Reader(s)

Kapfhammer, Gregory